This was my reading too. Interesting idea, but it took 10 pages of fluff to get to it and I didn't even believe the final idea when we got there. I started off reading the first part and thought he would get to the part where he realized he was managing context wrong. Never got there, instead he thought it was about the shape of the prompt.

Yes, this was super annoying to read. It was some core ideas and it was expanded into a way too long essay that boiled down to this guy doesn't know how to run agents.

Should he be harassed and deported for this?

Yes. If you're a guest in a country and don't follow the law you should be deported. You don't need waste money putting them on trial (except for murder/rape/etc), just deport.

How do you find out if someone followed the law without a trial?

That only works if the accused has committed a crime in their home country too.

Top of the line models still routinely get this wrong. This is just "make no mistakes" with more steps.

That is my understanding as well, but I see it so often and stuff changes so quickly it's hard to tell what is real.

There are a lot of missing parts to the story. If we assume the author left out everything that made them look bad, and including only what makes them look good then the result is a very incomplete feeling article.

For example: they asked for guidance and then the very next thing is them being fired. How did they respond to the coworker? Something is off here - the coworker who had messaged him about non-work topics TWO days in a row - then immediately reported him for his reply. What?

I was the one who let HR know that I sent her the message.

It definitely is. I got partway through and then realized it just keeps going without a point. Classic hallmark of AI writing.

You can't control the incentives, you can only control your reactions to those incentives.

We can control the incentives!

The idea that we should completely cede the environment and public space to corporations and give them total autonomy over the incentives that surround us is crazy.

My apologies. I was responding to the fact that animals and humans have base instincts that steer us towards bad food choices or over-stimulations. I was considering those instincts as the incentives that can't be changed.

On a macro scale, yes I agree that we should revolt against campaigns designed to exploit our animal nature. However on a personal basis, I do believe we owe it to ourselves to take radical responsibility over our own reactions to those things.

I agree. We have individual agency and also the corporations and systems around us have agency in ways that affect us.

My initial comment was disagreeing with the claim that any systemic problem foisted on the world is solely the fault of our individual wills to resist it.

Claude code can output Excalidraw format files which can be imported directly into the webapp. You can MCP it too if you want.

People would struggle to review code in a completely unfamiliar domain or part of the stack even before LLMs.

That’s why you need to write code to learn it.

No-one has ever learned skill just by reading/observing

what skill do they have? making stars? no they are skilled at observing, which is what they do.

I think understanding stellar processes and then using that understanding to theorize about other observations is a skill. My point was that observing can be a fantastic way to build a skill.. not all skills, but certainly some skills. Learning itself is as much an observation as a practice.

How would you find yourself in that situation before AI?

No, because they wouldn't be so foolish as to try it.

I noticed that too, I had to re-read it a few times to make sure I was understanding it properly, it seemed so out of place. Then I read the rest of the article and realized this guy has zero perspective on anything.